CN101949832B - Fourier transform infrared spectrum distortion identifying and processing method - Google Patents
Fourier transform infrared spectrum distortion identifying and processing method Download PDFInfo
- Publication number
- CN101949832B CN101949832B CN 201010268039 CN201010268039A CN101949832B CN 101949832 B CN101949832 B CN 101949832B CN 201010268039 CN201010268039 CN 201010268039 CN 201010268039 A CN201010268039 A CN 201010268039A CN 101949832 B CN101949832 B CN 101949832B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- gas
- absorptance
- formula
- concentration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a Fourier transform infrared spectrum distortion identifying and processing method. Aiming at the problem of spectrum distortion existed in a gas Fourier transform infrared spectrum online analysis process, the method comprises the following steps of: aiming at a specific application situation, determining possible components in the gas to be analyzed, searching spectral lines nearly insensitive to all the components in the whole spectrum wave number range according to the sample gas spectrum of the gas components, and correcting base line rule distortion in the real spectrum by adopting a spectrogram sectioned rotation and translation method according to the spectrum values of the spectral lines; performing gas concentration quantitative analysis by using the corrected spectrum, and performing spectrum reconstruction to the analysis result; and judging whether the spectrum has local irregular distortion according to the difference between the reconstructed spectrum and the actual spectrum, if so, abandoning the analysis result, and if the local irregular distortion is continuously caused, rescanning the background to eliminate the deviation of the spectrum distortion.
Description
Technical field
The present invention relates to gas on-line analysis field based on spectral analysis; Comprise the oil-gas component on-line analysis of oil, gas prospecting; The on-line analysis of electrical equipment fault gas; The mine safety gas on-line analysis, the on-line analysis of the factory smoke in the environmental protection, petrochemical complex associated gas, exhaust gases of internal combustion engines, the on-line monitoring of factory, naval vessels air quality.
Background technology
Spectral analysis is one of important method of gas composition and concentration analysis thereof.In recent years, along with the raising of data processing, data transmission, data storage technology, spectral analysis begins to be applied to the on-the-spot on-line analysis of gas.Because the parts that relate in the spectrometer are more; And many parts receive Effect of Environmental; The TGS detector temperature influence in the spectrometer for example, to such an extent as to spectrometer after long time continuous working, spectrum is prone to distortion; This causes analysis result bigger deviation possibly occur, badly influences applying of spectrometer.Though regularly carry out the probability that background scans helps to reduce to occur spectrum distortion, and the degree of distortion, owing to need clean air chamber when carrying out background scans, the time that this need spend several minutes, cause the loss of data of on-line analysis.Therefore, need carry out real-time distortion identification to spectrum, and different processing is made in different distortion, with under the prerequisite that guarantees accuracy of analysis, obliterated data as few as possible.
Summary of the invention
The objective of the invention is to; Be prone to disadvantage of distorted to spectrometer spectrum after long time continuous working; Adopt the identification and the treatment technology of spectrum distortion, different distortion types are made corresponding processing, with on the basis that guarantees the spectral analysis accuracy; Reduce the number of times of scanning background as far as possible, data loses in the minimizing gas on-line analysis process.
In order to realize above-mentioned task, the present invention adopts following technical solution:
At first to the definite gas composition that possibly exist of certain applications; And in the overall optical spectral limit, search insensitive to gas composition is arranged more; Or the low-down spectral line section of sensitivity, adopt the method for spectrogram segmentation rotation and translation to proofread and correct the regular distortion of spectrum then.On the basis of correction rule distortion, carry out conventional gas concentration spectral analysis, and carry out spectrum reconstruct according to analysis result.Relatively reconstruct spectrogram and actual light spectrogram judge according to both difference whether spectrum local irregularities's distortion takes place.If background scans is then carried out in the generation fortuitous distortion again,, obtain correct analysis result to obtain correct spectrum.Mainly comprise the steps:
1) to the definite gas composition that possibly exist of certain applications; Concentration is identical as far as possible; And each component gas spectrum that concentration is bigger is drawn in same figure, observes each component spectra in the overall optical spectral limit, searches insensitive to all gas component; Or the low-down spectral line section of sensitivity, be referred to as inert zone.The spectral line section low to the sensitivity degree, in the concentration range of each gas, according to linearization process, the sensitivity of estimation all gases.For the sensitivity S of i kind gas in j non-sensitive district
Ij, available formula (1) is confirmed:
s
ij＝(1-mean(v
ij))/c
i (1)
V in the formula
IjRepresent i kind gas several continuous spectrum values in j non-sensitive district.The quantity of continuous spectrum value is chosen between 3～20; Mean () explains the computing of averaging; c
iThe concentration of representing i kind gas one-component sample.
For the sensitivity S of i kind gas in j non-sensitive district
Ij, also available formula (2) is confirmed:
s
ij＝inv(Y
i*Y
i′)*Y
i*(1-V
ij)′ (2)
Y in the formula
iThe concentration value vector of representing i kind gas one-component sample; Y
i' represent vectorial Y
iTransposition; V
Ij=[mean (v
Ij1), mean (v
Ij2) ..., mean (v
IjN)] the spectral line mean vector in j non-sensitive district in the expression i kind gas one-component sample spectrum, mean (v
Ijk) (k=1,2 ..., the N) average of some spectral lines in j non-sensitive district of k one-component sample of expression i kind gas spectrum; Inv () representing matrix inversion operation.
2) between all adjacent two non-sensitive districts, translation and rotation are carried out in spectrum segment, carried out baseline correction.For the spectrum between two adjacent non-sensitive districts, can adopt formula (3) to realize:
Baseline
j＝1-mean(v
j)-S
jC′ (3a)
Baseline
j+1＝1-mean(v
j+1)-S
j+1C′ (3b)
Baserate
j＝(Baseline
j+1-Baseline
j)/(num
j+1-num
j) (3c)
v
j，j+1(num
j:num
j+1)＝v
j，j+1(num
j:num
j+1)+Baseline
j+([(num
j:num
j+1)]′-num
j)×Baserate
j?(3d)
V in the formula
jContinuous some spectral line values of representing the non-sensitive district of j; S
j=[s
1j, s
2j..., s
Mj] represent that the M component gas is vectorial in the sensitivity in the non-sensitive district of j; C '=[c
1, c
2..., c
M] expression M component gas concentration vector; Num
jExplain the center spectral line sequence number in the non-sensitive district of j; v
J, j+1(num
j: num
J+1) spectral value between the center spectral line sequence number of center spectral line sequence number in the non-sensitive district of j+1 in the expression j non-sensitive district.
3) with the analytical model of demarcating the spectrum after proofreading and correct is analyzed, obtained gas composition and concentration thereof.If portion gas concentration is bigger, and the part inert zone has less sensitivity to these gases, then revise the spectroscopic data of this inert zone again, and repeating step 2) with 3), be tending towards 0 up to the difference of twice adjacent modified value of this step;
4) gas composition and the concentration reconstruct spectrum thereof that obtain by analysis.
For the reconstruct of spectrum, at first to estimate the absorptance of every kind of gas.For given spectrometer, because light path is invariable, therefore converting absorptance is equivalent to absorptance.The conversion absorptance is the product of actual absorptance and light path.For every kind of gas, if the one-component sample of variable concentrations is arranged, the gas concentration that the conversion absorptance of every spectral line is calculated according to step 3), and the one-component sample is estimated.Estimation approach comprises piece-wise linearization estimation, Polynomial Estimation, spline method estimation.
Estimate for piece-wise linearization, if the concentration calculated value of i kind gas is c
I, x, and c
I, j＜c
I, x≤c
I, j+1, j=1,2 ..., N-1 is c wherein
I, j, c
I, j+1Represent j and j+1 one-component concentration of specimens of i kind gas respectively, N representes i kind gas one-component sample size.K bar spectral line place then, the conversion absorbance of this kind gas is:
rate＝(c
i，x-c
i，j)/(c
i，j+1-c
i，j) (4a)
δ
ikx＝(1-rate)×δ
ik，j+rate×δ
ik，j+1 (4b)
δ in the formula
Ik, jAnd δ
Ik, j+1The j and the conversion absorptance of j+1 one-component sample that represent i kind gas respectively at k bar spectral line.The conversion absorptance can adopt formula (5) to calculate:
δ
ik，j＝-log(v
ikj)/c
i，j (5)
V in the formula
IkjRepresent the spectral value of j one-component sample of i kind gas, the computing of log () expression natural logarithm at k bar spectral line.
For Polynomial Estimation, if the one-component concentration of specimens of i kind gas vector is C
i=[c
I, 1, c
I, 2..., c
I, N], the spectral line value vector of corresponding k bar spectral line is V
Ik=[v
Ik1, v
Ik2..., v
IkN], can obtain vectorial Δ=[δ at first with the conversion absorptance of formula (5) calculating sample
Ik, 1, δ
Ik, 2..., δ
Ik, N]
T, adopt formula (6) to convert the Polynomial Estimation of absorptance then:
δ
ikx＝P×B (6)
P=[log in the formula
M(c
I, x), log
M-1(c
I, x) ..., 1], B=[b
M, b
M-1..., b
0]
T, M＜N-1, and M>=1.B is confirmed by formula (7):
B＝inv(Q
TQ)×Q
T×Δ (7)
In the formula
Estimate for spline interpolation, directly under the Matlab environment, call spline () function and estimate:
δ
ikx＝spline(C
i，Δ，c
i，x) (8)
Perhaps at first adopt formula (9a) to obtain structure xp, and preserve
xp＝spline(C
i，Δ) (9a)
Convert absorptance by this structure and i kind gas concentration value with the ppval () function calculation under the Matlab environment then:
δ
ikx＝ppval(xp，c
i，x) (9b)
Consider within the specific limits in gas concentration; Its absorptance is desirable; Just almost invariable, in order to reduce the influence of estimated value of noise, in the estimation procedure of the conversion absorptance of each component gas, be threshold value T of every spectral line setting of every kind of gas to the conversion absorptance
Ik, when gas concentration is worth greater than this, with concentration T
IkThe time the conversion absorptance as the conversion absorptance of this kind gas, and need not adopt the described method of this step to calculate.T
IkMain consideration is set be among the i gas in the absorptance of k bar spectral line, and should the place spectrum smoothness.If absorptance is big, and there are absorption spike, then T
IkCorresponding value is smaller, otherwise, set the T of some greatly
IkValue.
After estimating to have got well the conversion absorptance of each component, then according to the Lamber-Beer theorem, the spectrum reconstruction value of k bar spectral line is calculated by formula (10):
Exp () expression nature exponent arithmetic in the formula.
In the process of spectrum reconstruct, for saving time, only consider that object gas and interference gas have the spectrum segment of absorption, other wave number section can not considered.
5) relatively reconstruct spectrum and actual spectrum, if the actual spectrum value of certain wave number section less than the reconstruct spectral value, and gap is bigger; Perhaps in the actual spectrum spectral value of certain wave number section greater than 1, and obvious with 1 gap, exceeded one spectral noise amplitude; Both difference opposite in sign of perhaps different wave number sections, and difference is bigger, explains that then possibly there is local distortion in spectrum; This time spectral analysis result error is bigger, must give up, and possibly rescan background.
Description of drawings
Fig. 1 (a) concentration is the Fourier transform middle infrared spectrum that is 1% methane, ethane, propane, isobutane, normal butane;
Fig. 1 (b) concentration is the Fourier transform middle infrared spectrum that is 1% isopentane, n-pentane, cyclopentane, isohexane and normal hexane;
The spectrum of baseline wander takes place in Fig. 2 (a);
Spectrum after Fig. 2 (b) baseline correction;
The spectrum and the reconstruct spectrum thereof of local distortion takes place in Fig. 3 (a) and Fig. 3 (b);
The normal spectrum and the reconstruct spectrum thereof of baseline wander only takes place in Fig. 3 (c).
Embodiment
Do not lose a characteristic of stock, the present invention is an example with the spectral analysis of the gas of required analysis in oil, the gas prospecting gas detection logging process, and embodiment of the present invention is described.The spectrometer that is wherein adopted is the T27 series Fourier transform mid infrared absorption spectrum appearance of Bruker company, and the spectrum wavenumber resolution is 4 wave numbers, and the spectral line value is a transmissivity, and each spectrogram has 1866 data.Specifically comprise the following steps:
1) search in gas componant and non-sensitive district
During oil, gas prospecting gas detection logging were used, the principal ingredient of gas was methane, ethane, propane, isobutane, normal butane, isopentane, n-pentane, carbon dioxide.Since the main composition of analyzing is light alkane; And the composition principal ingredient of oil is a heavy paraffin hydrocarbon, like octane etc., just possibly have cyclopentane, isohexane and normal hexane so in the gas detection logging process; Because more the alkane of high-order mainly is liquid at normal temperatures, gaseous state concentration is less.Therefore, in the mixed gas in the gas detection logging process, mainly comprise methane, ethane, propane, isobutane, normal butane, isopentane, n-pentane, carbon dioxide, cyclopentane, isohexane and normal hexane.Concentration is the Fourier transform mid infrared absorption spectrum of these 10 kinds of paraffin gas of 1% shown in accompanying drawing 1.Can find out by accompanying drawing 1, wave number 600,1100,2000,2500,3400 etc. locate near, the sensitivity of above-mentioned 11 kinds of gases is very little, therefore can be used as non-sensitive district.
For every kind of one-component sample; Owing to be to complete within a short period of time; Spectrum can not distort under one situation, and at most therefore the just faint translation of baseline need not do any processing yet; Directly average near above-mentioned five the wave number sections to every kind of one-component gaseous spectrum, and the least square method of employing formula (2) is estimated its sensitivity.For example, for n-pentane, one has 13 one-component samples.Near wave number 1100, each sample is got 5 spectral values continuously, and ask its average, obtain:
X＝[0.9998?0.9982?0.9981?1.0008?0.9999?0.9993?0.9967?0.9981?0.9973?0.9965?0.9960?0.99300.9903] (11)
Corresponding concentration is:
Y＝[0.01?0.02?0.05?0.1?0.2?0.5?1.0?2.0?3.0?4.0?5.0?7.0?10.0] (12)
Can find out that by formula (11) near the spectral line the wave number 1100 is very little to the sensitivity of n-pentane, therefore it made linearization process, and can not bring obvious variation to spectrum.So can in the hope of
S＝inv(Y*Y’)*Y*(1-X)′＝0.00096 (13)
Inv () representing matrix is inverted in the formula, (1-X) ' and the transposition of representing matrix (1-X).Use the same method and can also try to achieve of the sensitivity of 10 kinds of gases, obtain 11 kinds of gases at last and be at the sensitivity coefficient of this wave number section in this wave number section:
S
1100＝[0.00003?0.000096?0.00055?0.00044?0.00072?0.00061?0.00096?0.00042?0.000530.000850] (14)
Can also calculate near above-mentioned 11 kinds of gas sensitivity wave number 600 is:
S
600＝[0 0.000042 0.000076 0 0 0 0 0 0 0 0.00112] (15)
Near wave number 2000 sensitivity is:
S
2000＝[0.000022?0.00015?0.00033?0.00051?0.00075?0.00112?0.00071?0.00032?0.000920.00072?0.00002] (16)
Near 2500 wave numbers sensitivity is:
S
2500＝[0.000024?0.00017?0.00035?0.00061?0.00077?0.00132?0.00072?0.00033?0.000940.00075?0.00003] (17)
Near 3400 wave numbers sensitivity is:
S
3400＝[0.000006?0.00013?0.00028?0.0003?0.00042?0.00049?0.00053?0.00052?0.00051?0.00054?0.00011] (18)
2) between all adjacent two non-sensitive districts, translation and rotation are carried out in spectrum segment, carried out baseline correction;
For 5 non-sensitive districts in the step 1), can whole spectrogram be divided into 4 intervals, translation and rotation are carried out in each interval, i.e. the regular distortion of rectifiable spectrum.Because each spectrogram is made up of two row, first row are wave number values, and secondary series is and the corresponding spectral line value of first row that pairing spectral line sequence number is followed successively by 273,753,1023,1503,1758 near 3400,2500,2000,1100 and 600 wave numbers.So get 5 spectral line values and ask mean deviation in every section non-sensitive district, suppose that the concentration vector of 11 kinds of gases to be analyzed is C, so, according to formula (3), can adopt following source code to realize first spectral shift and rotation:
Baseline3400=1-mean (data (271:275,2))-S
3400C '; % asks the deviation at wave number 3400 places
Baseline2500=1-mean (data (751:755,2))-S
2500C '; % asks the deviation at wave number 2500 places
Baserate2500=(Baseline2500-Baseline3400)/(753-273); % asks the slope between the wave number 2500 to 3400
Data (1:753,2)=data (1:753,2)+Baseline3400+ ([1:753] '-271) * Baserate2500; % corrects the spectrum between the wave number 2500 to 3400
Baseline2000=1-mean (data (1021:1025,2))-S
2000C '; % asks the deviation at wave number 2000 places
Baserate2000=(Baserate2000-Baserate2500)/(1023-753); % asks the slope between the wave number 2000 to 2500
Data (754:1023,2)=data (754:1023,2)+Baseline2500+ ([754:1023] '-753) * Baserate2000; % corrects the spectrum between the wave number 2500 to 3400
Baseline1100=1-mean (data (1501:1505,2))-S
1100C '; % asks the deviation at wave number 1100 places
Baserate1100=(Baseline1100-Baseline2000)/(1503-1023); % asks the slope between the wave number 1100 to 2000
Data (1024:1503,2)=data (1024:1503,2)+Baseline2000+ ([1024:1503] '-1023) * Baserate1100; % corrects the spectrum between the wave number 1100 to 2000
Baseline600=1-mean (data (1756:1760,2))-S
600C '; % asks the deviation at wave number 1100 places
Baserate600=(Baseline600-Baseline1100)/(1758-1503); % asks the slope between the wave number 600 to 1100
Data (1504:1866,2)=data (1504:1866,2)+Baseline1100+ ([1504:1866] '-1503) * Baserate600; % corrects the spectrum between the wave number 600 to 1100
In continuous on-line spectral analysis process, the gas concentration vector C in the above-mentioned source code is set at last analysis result.If analyze for the first time, then it is set to 0.For the accompanying drawing in this embodiment 2 (a), three spectrogram: data1, data2 and data3 are arranged.Comparative drawings figs 2 (a) and accompanying drawing 1 can be known; Since wave number does not have strong absorption peak near 2900 places, explain that the concentration of various alkane is very little in these three gases that spectrogram characterized; Therefore; In wave number 800 to wave number 1100, and in 2500 wave numbers in 3400 wave-number ranges, almost be that an amplitude is 1 straight line.But in the accompanying drawing 2 (a); Wave number 800 to the spectral value of the spectrum segment of wave number 1100 obviously greater than 1; And inclination is arranged slightly, and being significantly less than 1 in 2500 wave numbers to the spectrum segment in 3400 wave-number ranges, its inclined degree is bigger slightly to the spectrum segment of wave number 1100 than wave number 800; Therefore there is the baseline regular distortion, need proofreaies and correct.It is 0 that gas concentration vector C is set, and the spectrogram that obtains after adopting the source code of this step to proofread and correct is shown in accompanying drawing 2 (b);
3) with the analytical model of demarcating the spectrum after proofreading and correct is analyzed, obtained gas composition and concentration thereof.The analytical model of demarcating comprises it being polynomial expression, neural network, SVMs.For example, for the methane in this embodiment, suppose that it exists certain polynomial expression analytical model to be:
conmet＝(log(data(1410，2))-log(data(1397，2)))*2.22-(log(data(1540，2))-log(data(1570，2)))*0.31-(log(data(1499，2))-log(data(1462，2)))*0.08 (19a)
C
methane＝0.26*conmet^2+conmet*0.9 (19b)
Then set step 2) in C be 0 vector; And after accompanying drawing 2 (a) spectrum proofreaied and correct; The methane concentration that calculating formula (19) can be tried to achieve among this spectrogram data1, data2 and the data3 is respectively respectively: 0.0196,0.0071 and 0.0473; Promptly 0.0196%, 0.0071% and 0.0473%, 196ppm, 71ppm and 473ppm in other words.Through analyzing, data1, data2 and the corresponding gas concentration vector of three spectrograms of data3 are:
C
1＝[0.0196?0.0032?0.0017 0 0 0 0 0 0 0 0.3572] (20a)
C
2＝[0.0071?0.0012?0.0003 0 0 0 0 0 0 0 0.2956] (20b)
C
3＝[0.0473?0.0021?0.0023 0 0 0 0 0 0 0 0.4763] (20c)
If portion gas concentration is bigger, then use the gas concentration value step of replacing 2 of latest computed) in gas concentration vector C, and repeating step 2) with 3), up to step 2) in the difference of adjacent twice deviate in any one non-sensitive district less than certain threshold value.The threshold value in each non-sensitive district can be set at the noise amplitude of this wave number section spectrum.Because each component gas concentration is very little, therefore by step 2) proofread and correct and once get final product.
4) gas composition and the concentration reconstruct spectrum thereof that obtain by analysis
For simplicity, in this instance, wave number section 2800～3200, the threshold value of each component paraffin gas all is set to 0.1; And in wave number section 700～1300, the threshold value of each component paraffin gas all is set to 1.Because each component gas concentration is all less than 0.1, so directly adopt the corresponding conversion absorptance of threshold value to get final product.That is to say that in wave number section 700～1300, it is 0.1% o'clock conversion absorptance that the conversion absorptance of each component is directly got concentration, and in wave number section 2800～3200, it is 1% o'clock conversion absorptance that the conversion absorptance of each component is directly got concentration.For example; N-pentane is No. 7 component in gas composition, can be known by formula (12), and 0.1% concentration is arranged the 4th in the one-component sample of n-pentane; The spectral line value that n-pentane is located at the 579th spectral line (corresponding to wave number 2881.5) is 0.9462, so its conversion absorptance is calculated by formula (5):
δ
7，579，4＝-log(0.9462)/0.1＝0.5530
By that analogy, can try to achieve 11 component gas in the conversion absorptance of 579 spectral lines and constitute vectorial Δ and obtain:
Δ＝[0.0200?0.1410?0.5393?0.5402?0.5221?0.7526?0.5530?0.5493?0.7439?0.5526?0.0002]?(21)
Wushu (20) and (21) substitution formula (10) can be tried to achieve the reconstruct spectral value that data1, data2 and three spectrograms of data3 go out at the 579th spectral line respectively:
v
1，579＝exp(-C
1Δ′)＝0.9982
v
2，579＝exp(-C
2Δ′)＝0.9995
v
3，579＝exp(-C
3Δ′)＝0.9974
Likewise, the gas concentration vector C that tries to achieve in the step 3)
1, C
2And C
3, and every spectral line conversion absorptance δ
IkxSubstitution formula (10), then can try to achieve the reconstruct spectral line value of every spectral line.So, for the data1 in the accompanying drawing 2, data2 and data3, proofread and correct forward and backwardly, and the spectrum of reconstruct is respectively shown in accompanying drawing 3 (a), 3 (b) and 3 (c).
5) reconstruct spectrum and the actual spectrum in the comparative drawings figs 3 (a) can find that near wave number 1060, revised spectral value obviously exceeds 1.0020; In fact, the noise amplitude at this place has only about 0.0015, so possibly there is local distortion in data1 spectrum; This time analysis result deviation maybe be bigger; As possible, need rescan background, to obtain analysis result preferably; Reconstruct spectrum and actual spectrum in the comparative drawings figs 3 (b) can be found equally again, and near wave number 1045, revised spectral value has reached 1.0020, so possibly also there is local distortion in data2 spectrum.And can find by accompanying drawing 3 (c), revised spectral value maximum only about 1.0010, so data3 is the good spectrum that local distortion does not take place, and with peg model this spectrum analyzed, its confidence level is very high.In fact, (b) can find out by accompanying drawing 2, and in wave number 1000 to 1200 scopes, the spectrum of data1 and data2 is the ribbed that frequency does not wait, and this itself is exactly a kind of performance of spectrum local distortion.
Above content is to combine concrete preferred implementation to further explain that the present invention did; Can not assert that embodiment of the present invention only limits to this; Those of ordinary skill for technical field under the present invention; Under the prerequisite that does not break away from the present invention's design, can also make some simple deduction or replace, all should be regarded as belonging to the present invention and confirm scope of patent protection by claims of being submitted to.
Claims (5)
1. a Fourier transform infrared spectroscopy distortion is discerned and disposal route; It is characterized in that: at first confirm that to the application scenario M that possibly exist organizes gas composition; In the overall optical spectral limit, search the insensitive spectral line of all gas component; And continuous insensitive spectral line is divided into a non-sensitive district, mark off T non-sensitive district altogether; After having confirmed T non-sensitive district, calculate of the sensitivity of each component gas in each non-sensitive district, adopt the regular distortion of the method correction spectrum of spectrogram segmentation rotation and translation; On the basis of correction rule distortion, carry out conventional gas concentration spectral analysis, and carry out spectrum reconstruct according to analysis result; Relatively reconstruct spectrogram and actual light spectrogram judge according to both difference whether spectrum local irregularities's distortion takes place; If this time analysis result is then abandoned in the generation fortuitous distortion, the prompting mistake, or carry out background scans again, to obtain correct spectrum, obtain correct analysis result.
2. the method for claim 1 is characterized in that, saidly continuous insensitive spectral line is divided into a non-sensitive district adopts following mode:
Concentration is identical as far as possible, and each bigger component gas spectrum of concentration draws in same figure, observes each component spectra in the overall optical spectral limit, searches to all gas component the insensitive or low-down spectral line section of sensitivity, is referred to as inert zone; For the low spectral line section of sensitivity, in the concentration range of each gas, according to linearization process, the sensitivity of estimation all gases; For the sensitivity S of i kind gas in j non-sensitive district
Ij, replace with of the sensitivity of high concentration value sample in this district, perhaps estimate with the least square method method; For the former, confirm with formula (1):
s
ij＝(1-mean(v
ij))/c
i (1)
V in the formula
IjRepresent i kind gas several continuous spectrum values in j non-sensitive district; The quantity of continuous spectrum value is chosen between 3～20; Mean () explains the computing of averaging; c
iRepresent the highest concentration in the i kind gas one-component sample;
The least square method method estimates that available formula (2) is definite:
s
ij＝inv(Y
i*Y
i’)*Y
i*(1-V
ij)′ (2)
Y in the formula
iThe concentration value vector of representing i kind gas one-component sample; V
Ij=[mean (v
Ij1), mean (v
Ij2) ..., mean (v
IjN)] the spectral line mean vector in j non-sensitive district in the expression i kind gas one-component sample spectrum, mean (v
Ijk) (k=1,2 ..., the N) average of some spectral lines in j non-sensitive district of k one-component sample of expression i kind gas spectrum.
3. the method for claim 1 is characterized in that, following method is adopted in the correction of the regular distortion in the said actual light spectrogram:
Baseline
j＝1-mean(v
j)-S
jC′ (3a)
Baseline
j+1＝1-mean(v
j+1)-S
j+1C′ (3b)
Baserate
j＝(Baseline
j+1-Baseline
j)/(num
j+1-num
j) (3c)
v
j，j+1(num
j:num
j+1)＝v
j，j+1(num
j:num
j+1)+Baseline
j+([(num
j:num
j+1)]′-num
j)×Baserate
j
(3d)
V in the formula
jContinuous some spectral line values of representing the non-sensitive district of j; S
j=[s
1j, s
2j..., s
Mj] represent that the M component gas is vectorial in the sensitivity in the non-sensitive district of j; C '=[c
1, c
2..., c
M] expression M component gas concentration vector; Num
jExplain the center spectral line sequence number in the non-sensitive district of j; v
J, j+1(num
j: num
J+1) spectral value between the center spectral line sequence number of center spectral line sequence number in the non-sensitive district of j+1 in the expression j non-sensitive district.
4. the method for claim 1 is characterized in that, the reconstruct of said spectrum adopts following steps to realize:
At first estimate the conversion absorptance of every kind of gas; The conversion absorptance is the product of actual absorptance and light path, and for every kind of gas, if the one-component sample of variable concentrations is arranged, the conversion absorptance of every spectral line is according to gas concentration, and the one-component sample is estimated; Estimation approach is the piece-wise linearization estimation technique, Polynomial Estimation method or the spline method estimation technique;
Estimate for piece-wise linearization, if the concentration calculated value of i kind gas is c
I, x, and c
I, j＜c
I, x≤c
I, j+1, j=1,2 ..., N-1 is c wherein
I, j, c
I, j+1Represent j and j+1 one-component concentration of specimens of i kind gas respectively, N representes i kind gas one-component sample size; K bar spectral line place then, the conversion absorbance of this kind gas is:
rate＝(c
i，x-c
i，j)/(c
i，j+1-c
i，j) (4a)
δ
ikx＝(1-rate)×δ
ik，j+rate×δ
ik，j+1 (4b)
δ in the formula
Ik, jAnd δ
Ik, j+1The j and the conversion absorptance of j+1 one-component sample that represent i kind gas respectively at k bar spectral line; The conversion absorptance adopts formula (9) to calculate:
δ
ik，j＝-log(v
ikj)/c
i，j (5)
V in the formula
IkjRepresent the spectral value of j one-component sample of i kind gas, the computing of log () expression natural logarithm at k bar spectral line;
For Polynomial Estimation, if the one-component concentration of specimens of i kind gas vector is C
i=[c
I, 1, c
I, 2..., c
I, N], the spectral line value vector of corresponding k bar spectral line is V
Ik=[v
Ik1, v
Ik2..., v
IkN], can obtain vectorial Δ=[δ at first with the conversion absorptance of formula (5) calculating sample
Ik, 1, δ
Ik, 2..., δ
Ik, N]
T, adopt formula (6) to convert the Polynomial Estimation of absorptance then:
δ
ikx＝P×B (6)
P=[log in the formula
M(c
I, x), log
M-1(c
I, x) ..., 1], B=[b
M, b
M-1..., b
0]
T, M＜N-1, and M>=1; B is confirmed by formula (7):
B＝inv(Q
TQ)×Q
T×Δ (7)
In the formula
Estimate for spline interpolation, directly under the Matlab environment, call spline () function and estimate:
δ
ikx＝spline(C
i，Δ，c
i，x) (8)
Perhaps at first adopt formula (9a) to obtain structure xp, and preserve
xp＝spline(C
i，Δ) (9a)
Convert absorptance by this structure and i kind gas concentration value with ppval () function calculation then:
δ
ikx＝ppval(xp，c
i，x) (9b)
In gas concentration within the specific limits; Its absorptance is desirable; Just almost invariable, in order to reduce the influence of estimated value of noise, in the estimation procedure of the conversion absorptance of each component gas, be threshold value T of every spectral line setting of every kind of gas to the conversion absorptance
Ik, when gas concentration is worth greater than this, with concentration T
IkThe time the conversion absorptance as the conversion absorptance of this kind gas, and needn't adopt the described method of this step to calculate; T
IkMain consideration is set be among the i gas in the absorptance of k bar spectral line, and should the place spectrum smoothness; If absorptance is big, and there are absorption spike, T
IkCorresponding value is smaller, otherwise, set big;
After estimating to have got well the conversion absorptance of each component, then according to the Lamber-Beer theorem, the spectrum reconstruction value of k bar spectral line is calculated by formula (10):
Exp () expression nature exponent arithmetic in the formula.
5. the method for claim 1 is characterized in that, the identification of said local irregularities distortion adopts following method to carry out with processing:
Compare reconstruct spectrum and actual spectrum; If the actual spectrum value of certain wave number section is greater than the reconstruct spectral value, and gap is bigger, perhaps in the actual spectrum spectral value of certain wave number section greater than 1 and obvious with 1 gap; Exceeded general spectral noise amplitude; Explain that then possibly there is local irregularities's distortion in spectrum, this time spectral analysis result error is bigger, must give up; If occur local irregularities's distortion continuously, then need rescan background.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010268039 CN101949832B (en) | 2010-09-02 | 2010-09-02 | Fourier transform infrared spectrum distortion identifying and processing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010268039 CN101949832B (en) | 2010-09-02 | 2010-09-02 | Fourier transform infrared spectrum distortion identifying and processing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101949832A CN101949832A (en) | 2011-01-19 |
| CN101949832B true CN101949832B (en) | 2012-07-25 |
Family
ID=43453407
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010268039 Active CN101949832B (en) | 2010-09-02 | 2010-09-02 | Fourier transform infrared spectrum distortion identifying and processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101949832B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102562051B (en) * | 2011-12-31 | 2015-07-15 | 中国石油集团川庆钻探工程有限公司 | Method for processing original data of gas-measuring plate |
| US9182280B1 (en) * | 2014-08-08 | 2015-11-10 | Thermo Scientific Portable Analytical Instruments Inc. | Method for reducing frequency of taking background/reference spectra in FTIR or FTIR-ATR spectroscopy and handheld measurement device embodying same |
| WO2018170035A1 (en) * | 2017-03-14 | 2018-09-20 | Saudi Arabian Oil Company | Collaborative sensing and prediction of source rock properties |
| CN108169156B (en) * | 2017-12-08 | 2020-05-01 | 中国矿业大学 | Three-stage correction method for Fourier transform infrared spectrum in-situ diffuse reflectance spectrogram |
| CN108287137A (en) * | 2017-12-22 | 2018-07-17 | 必欧瀚生物技术（合肥）有限公司 | A kind of baseline correction method based on piecewise polynomial fitting |
| CN109859137B (en) * | 2019-02-14 | 2023-02-17 | 重庆邮电大学 | Wide-angle camera irregular distortion global correction method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004163311A (en) * | 2002-11-14 | 2004-06-10 | Konica Minolta Holdings Inc | Infrared absorption spectrum measuring method and device |
| CN1664851A (en) * | 2005-04-14 | 2005-09-07 | 中国国土资源航空物探遥感中心 | Imaging spectrum data processing system and imaging spectrum data processing method |
| WO2005103945A2 (en) * | 2004-03-31 | 2005-11-03 | Sensys Medical, Inc. | Method and apparatus for enhanced estimation of an analyte property through multiple region transformation |
| US20060197957A1 (en) * | 2005-03-07 | 2006-09-07 | Jones Christopher M | Method to reduce background noise in a spectrum |
| CN101271063A (en) * | 2008-05-07 | 2008-09-24 | 陕西恒智科技发展有限公司 | Process of vector machine emendation model method supported by gas infrared spectrum analysis |
| CN101694466A (en) * | 2009-10-19 | 2010-04-14 | 浙江大学 | Correction method of etalon effect of back lighting type CCD array spectrometer |
-
2010
- 2010-09-02 CN CN 201010268039 patent/CN101949832B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004163311A (en) * | 2002-11-14 | 2004-06-10 | Konica Minolta Holdings Inc | Infrared absorption spectrum measuring method and device |
| WO2005103945A2 (en) * | 2004-03-31 | 2005-11-03 | Sensys Medical, Inc. | Method and apparatus for enhanced estimation of an analyte property through multiple region transformation |
| US20060197957A1 (en) * | 2005-03-07 | 2006-09-07 | Jones Christopher M | Method to reduce background noise in a spectrum |
| CN1664851A (en) * | 2005-04-14 | 2005-09-07 | 中国国土资源航空物探遥感中心 | Imaging spectrum data processing system and imaging spectrum data processing method |
| CN101271063A (en) * | 2008-05-07 | 2008-09-24 | 陕西恒智科技发展有限公司 | Process of vector machine emendation model method supported by gas infrared spectrum analysis |
| CN101694466A (en) * | 2009-10-19 | 2010-04-14 | 浙江大学 | Correction method of etalon effect of back lighting type CCD array spectrometer |
Non-Patent Citations (1)
| Title |
|---|
| 郭小方 等.机载成像光谱图象边缘辐射畸变校正.《中国图象图形学报》.2000,第5卷(第1期),16-20. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101949832A (en) | 2011-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101949832B (en) | Fourier transform infrared spectrum distortion identifying and processing method | |
| Correia et al. | Portable near infrared spectroscopy applied to fuel quality control | |
| US8735820B2 (en) | Direct match spectrographic determination of fuel properties | |
| Corsetti et al. | Characterization of gasoline/ethanol blends by infrared and excess infrared spectroscopy | |
| Abbas et al. | PLS regression on spectroscopic data for the prediction of crude oil quality: API gravity and aliphatic/aromatic ratio | |
| US10942117B2 (en) | Estimation of water interference for spectral correction | |
| US8481942B2 (en) | Direct match spectrographic determination of fuel properties | |
| JP2005512051A (en) | A method of analyzing unknowns as a blend of known substances calculated to match specific analytical data and predicting the properties of unknowns based on the calculated blend | |
| Li et al. | Real-time infrared gas detection based on an adaptive Savitzky–Golay algorithm | |
| DK178302B1 (en) | A computer assisted method for quantification of total hydrocarbon concentrations and pollution type apportionment in soil samples by use of gc-fid chromatograms | |
| CN103364364B (en) | Crude oil property rapid detection method based on recombination prediction technology | |
| Choquette et al. | Identification and quantitation of oxygenates in gasoline ampules using Fourier transform near-infrared and Fourier transform Raman spectroscopy | |
| Marinović et al. | Prediction of diesel fuel properties by vibrational spectroscopy using multivariate analysis | |
| JP2017523400A (en) | System, method, and apparatus for monitoring an optical hydrocarbon gas composition | |
| Barthès et al. | Improvement in spectral library-based quantification of soil properties using representative spiking and local calibration–The case of soil inorganic carbon prediction by mid-infrared spectroscopy | |
| Yu et al. | A novel integrated approach to characterization of petroleum naphtha properties from near-infrared spectroscopy | |
| Kuzmiakova et al. | An automated baseline correction protocol for infrared spectra of atmospheric aerosols collected on polytetrafluoroethylene (Teflon) filters | |
| Garrigues et al. | Multivariate calibrations in Fourier transform infrared spectrometry for prediction of kerosene properties | |
| Li et al. | Selection of the effective characteristic spectra based on the chemical structure and its application in rapid analysis of ethanol content in gasoline | |
| Šašić et al. | Comparing the predictions by NIR spectroscopy based multivariate models for distillation fractions of crude oils by F-test | |
| Sasic et al. | Multivariate calibration of total acid number in crude oils via near-infrared spectra | |
| de Souza et al. | PAT soft sensors for wide range prediction of key properties of diesel fuels and blending components for the oil industry | |
| Kartnaller et al. | Evaluating the kinetics of the esterification of oleic acid with homo and heterogeneous catalysts using in-line real-time infrared spectroscopy and partial least squares calibration | |
| CN106970042B (en) | Method for detecting impurity and moisture content of carrageenin | |
| Laurentino Alves | Vibrational spectroscopy for the quantification of hydrotreated vegetable oil (hvo) advanced biofuels in petroleum-derived fuel blends: a minireview |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |